This is a fully remote position, open to applicants in United States.

📋 Description

• Generate innovative concepts for challenging machine learning engineering tasks within a long-term AI research initiative at a premier AI laboratory.

• Design tasks that are complex enough to pose a challenge to cutting-edge models, typically solvable in no more than 2 out of 8 attempts.

• Prioritize inference, evaluation, and systems work over intensive model training.

• Develop a combination of tasks that utilize both GPU and CPU resources.

• Clearly articulate the reasons each problem presents difficulty for a frontier model.

• Primarily contribute ideas; constructing or reviewing tasks is not a requirement.

• Engage in projects focused on training and enhancing advanced AI systems.

⛳️ Requirements

• Current or recent position as an ML research engineer, research scientist, or applied scientist.

• Practical experience in training, fine-tuning, or optimizing models using PyTorch or JAX.

• In-depth understanding of at least one specialized area: LLM pretraining, alignment or evaluations; GPU inference and performance; reinforcement learning; or scientific ML.

• Capability to explain why a particular problem is challenging for a frontier model.

• Experience with benchmark or evaluation inquiries is advantageous.

• A PhD in ML, CS, or physics, or an MSc with a strong research output is preferable.

• Publications are a beneficial addition.

• Previous contracts with Mercor demonstrating strong performance are a plus.

• Must not be an H-1B or STEM OPT candidate.

• Independent contractor engagement is required.